research paper series
Globalisation and Labour Markets Programme
Research Paper 2001/25
Foreign Direct Investment, Agglomerations and
Demonstration Effects: An Empirical Investigation
By F. Barry, H. Görg and E. Strobl
The Centre acknowledges financial support from The Leverhulme Trust
under Programme Grant F114/BF
The Authors
Frank Barry is a Senior Lecturer in the Department of Economics, University College Dublin.
Holger Görg is a Research Fellows in The Leverhulme Centre for Research on Globalisation
and Economic Policy, University of Nottingham and Eric Strobl is a Lecturer in the Department
of Economics, University College Dublin.
Acknowledgements
We are grateful to participants of a CEPR network meeting in Madrid and a seminar at the
University of Manchester for helpful comments on an earlier draft. All remaining errors are, of
course, our own.
Financial support through the European Commission Fifth Framework
Programme (Grant No. HPSE-CT-1999-00017) and the Leverhulme Trust (Grant No. F114/BF)
is gratefully acknowledged.
Foreign Direct Investment, Agglomerations and Demonstration Effects:
An Empirical Investigation
by
F. Barry, H. Görg and E. Strobl
Abstract
Many previous studies have shown that the localisation of firms can be an important factor in
attracting new foreign direct investment into a host country. What has been missing in this
literature thus far, however, is an investigation into the reasons why industry clusters attract
firms. We distinguish between “efficiency agglomerations” as firms locating close to each
other because they can increase their efficiency by doing so, and “demonstration effects”,
whereby existing firms send signals to new investors as to the reliability of the host country and
newly entering firms follow previous firms. In this paper we try to disentangle these two
effects, by examining the location of US and UK firms in Ireland. We calculate proxies for
“efficiency agglomerations” and “demonstration effects” and include these proxies in an
empirical model of the location decision of firms. For US firms, we find that both efficiency
agglomeration and demonstration effects are important determinants of entry. For UK firms,
however, the evidence is not as clear cut.
Outline
1. Introduction
2. Description of the Data
3. Econometric Model and Methodology
4. Econometric Results
5. Summary Conclusions
Non-Technical Summary
Economists have recognised the importance of agglomeration benefits for the location of firms for a long
time. The literature generally postulates three reasons for the emergence of agglomerations. Industrial
districts in which firms benefit from locating close to each other arise, it is argued, because of (i)
knowledge spillovers between firms, (ii) the advantages provided by thick markets in specialised factors,
in particular labour, and (iii) the scope for backward and forward linkages between customer and supplier
firms. If these conditions exist, firms can increase efficiency by locating close to other firms, leading to
agglomeration of industries. Another strand of literature, however, points out that even if these “efficiency
agglomeration” effects are not prevalent firms may find it rational to agglomerate spatially. If there is
uncertainty about locations in which to invest, investors may exhibit a tendency to imitate each others’
location decisions. This arises because investors locating in a “good” location provide a signal to other
investors which have an incentive to choose the same “good” location for their investment. This
mechanism is referred to as “demonstration effects”.
We conduct an empirical analysis of the factors that attract inward foreign direct investment into the Irish
economy, attempting to distinguish between efficiency agglomerations and demonstration effects. As
foreign firms face greater uncertainties than domestic firms in the host country, they may have strong
incentives to follow previous investors because of the signal they send as to the reliability of the host
country location. In other words, even if the “efficiency agglomeration” reasons for the development of
agglomerations are not important, firms may choose the same locations due to demonstration effects.
Ireland provides a useful test case to investigate the importance of efficiency agglomeration and
demonstration effects in attracting inward FDI for a number of reasons. First there is the strong presence
of foreign-owned multinational companies in the Irish economy; in 1998, foreign multinationals accounted
for 47 per cent of manufacturing employment and 82 per cent of net output. Second is the substantial
presence of foreign-owned firms in the high-tech sector allowing us to analyse the relative importance of
efficiency agglomeration and demonstration effects across sectors differentiated by degree of technology.
Third are the differences between firms from the US and the second most predominant source country,
the UK. US-owned firms exhibit much higher labour productivity than UK-owned firms, are much more
strongly export-oriented, and are much more likely to be located in high-tech sectors.
In order to investigate whether efficiency agglomeration and demonstration effects may have impacted on
inward FDI in Ireland we model empirically the location decision of foreign-owned firms. In particular, we
focus on firms from the US and the UK as we would expect these two nationality groups to behave
differently due to their different average characteristics as pointed out above. We calculate proxies for the
effect of efficiency agglomerations and demonstration effects, and include these in the model of foreign
firms’ location. We find that, for US firms, both efficiency agglomeration and demonstration effects are
important factors, while the evidence for UK firms is not as clear-cut.
1
1.
Introduction
Economists have recognised the importance of agglomeration benefits for the location of
firms for a long time, the standard reference being to Marshall (1920). The implications of
agglomerations have recently been analysed extensively in the growing “new economic
geography” literature, see, for example, Krugman and Venables (1995, 1996). Following
Marshall, the new economic geography literature postulates three reasons for the emergence
of agglomerations. Industrial districts in which firms benefit from locating close to each
other arise, it is argued, because of (i) knowledge spillovers between firms, (ii) the
advantages provided by thick markets in specialised factors, in particular labour, and (iii)
the scope for backward and forward linkages between customer and supplier firms.1 If
these conditions exist, firms can increase efficiency by locating close to other firms, leading
to agglomeration of industries.
DeCoster and Strange (1993), however, have pointed out that even if these efficiency
reasons are not prevalent firms may find it rational to agglomerate spatially. If there is
uncertainty about locations in which to invest, investors may exhibit a tendency to imitate
each others’ location decisions. This arises because investors locating in a “good” location
provide a signal to other investors, and to banks which provide the funds for investments.
Banks conclude that investments in good locations have higher probabilities of success and
provide funding for investments in good locations more forthcoming than for investments
in bad locations. As other firms are aware of this choice mechanism, they have an incentive
to choose the same “good” location for their investment.
Empirically it is, of course, difficult to distinguish between the “efficiency” factors leading
to agglomerations a la Marshall and New Economic Geography, and the “demonstration
effect” as discussed by DeCoster and Strange. Although there have been a few studies
investigating the importance of industrial agglomerations for the location of foreign direct
investment (FDI) in the host country (see Wheeler and Mody, 1992; Head et al., 1995,
2000; Braunerhjelm and Svensson, 1996, 1998; Barrell and Pain, 1999), these papers do not
attempt to assess the relative importance of efficiency agglomeration factors compared to
the demonstration effect.
1 See Ottaviano and Puga (1998) and Fujita and Thisse (1996) for a fuller discussion of the reasons for
agglomerations.
2
The papers by Head et al. (1995, 2000) are perhaps most closely related to our paper and,
therefore, deserve a more detailed description. Specifically, these papers examine the
location of Japanese firms across US states using data for the 1980s. The location choice of
Japanese firms is modelled by a conditional logit regression and includes a proxy for the
effect of the presence of Japanese firms already present in the location, which they refer to
as a proxy for “agglomerations”. This proxy is defined as the number of Japanese firms in
the sector and region. The estimation yields a positive coefficient for the variable which
Head et al. take as evidence for the importance of agglomeration economies for the location
of Japanese firms. They do not, however, attempt to distinguish the effect of demonstration
effects and efficiency agglomerations as discussed above.
In this paper we attempt to shed light on this issue. We conduct an empirical analysis of the
factors that attract inward foreign direct investment into the Irish economy, focusing
particularly on the importance of efficiency agglomerations and demonstration effects.
Krugman (1997) points out that as foreign firms face greater uncertainties than domestic
firms in the host country, they may have strong incentives to follow previous investors
because of the signal they send as to the reliability of the host country location. In other
words, even if the “efficiency” reasons for the development of agglomerations a la Marshall
are not important, firms may choose the same locations due to demonstration effects. Such
demonstration effects have been alluded to by both Krugman (1997) and Barry and Bradley
(1997) in discussing the strong growth in inward FDI into Ireland over the last 15 years or
so.
Barry and Bradley (1997) write that "surveys of executives of newly arriving companies in
the computer, instrument engineering, pharmaceutical and chemical sectors indicate that
their location decision is now strongly influenced by the fact that other key market players
are already located in Ireland" (p. 1804). Krugman (1997) argues that Ireland now enjoys
the advantage of being able to demonstrate its reliability as a host country to new investors.
In the electronics industry, for example, it is host to “twenty of the top twenty-five US hightechnology companies” (White, 2000, p. 290).2
According to Krugman (1997), this
demonstration effect attracts other firms to locate in Ireland in the vicinity of the sector's
leading firms.
2
Firms with operations in Ireland include Compaq, Dell, Digital Equipment, Gateway Computers, Hewlett
Packard, IBM, Intel, Microsoft and Netscape.
3
For a number of reasons Ireland provides a useful test case to investigate the importance of
efficiency agglomeration and demonstration effects in attracting inward FDI. First there is
the strong presence of foreign-owned multinational companies (MNCs) in the Irish
economy, which is evident from data taken from the Irish Census of Industrial Production:
In 1998, foreign multinationals accounted for 47 per cent of manufacturing employment, 82
per cent of net output and 88 per cent of manufacturing exports. Second is the substantial
presence of foreign-owned firms in the high-tech sector, as evident from the data presented
in Table 1, allowing us to analyse the relative importance of traditional agglomeration and
demonstration effects across sectors differentiated by degree of technology. Third are the
differences between firms from the US and the second most predominant source country,
the UK. Illustrative differences presented in Table 1 show that the US-owned firms exhibit
much higher labour productivity than UK-owned firms, are much more strongly exportoriented, and are much more likely to be located in high-tech sectors.3
[Table 1]
In order to investigate whether efficiency agglomeration and demonstration effects may
have impacted on inward FDI, we model empirically the location decision of foreign-owned
firms. In particular, we focus on firms from the US and the UK. We would expect these
two nationality groups to behave differently due to their different average characteristics as
pointed out above. We calculate proxies for the effect of efficiency agglomerations and
demonstration effects, and include these in the model of foreign firms’ location. We find
that, for US firms, both efficiency agglomeration and demonstration effects are important
factors, while the evidence for UK firms is not as clear-cut.
The remainder of the paper is structured as follows. In Section 2 we present the data used
for the empirical analysis. Section 3 outlines the econometric methodology while Section 4
presents the estimation results. Section 5 summarises the main results and concludes.
2.
Description of the Data
Our main data source for our analysis of the importance of efficiency agglomerations and
demonstration effects on entry of foreign firms in Ireland is the Forfás Employment Survey
of Irish manufacturing firms. This survey has been undertaken annually by Forfás, the
3
In line with material presented later in the paper, these broad sectors are: Chemicals and Pharmaceuticals,
4
policy and advisory board for industrial development in Ireland, since 1973. Data are
available to us up to 1996. The survey covers virtually all active manufacturing companies,
and the response rate is generally over 99 per cent, providing a sample of over 15,000 firms.
For these firms we have information on employment levels, nationality of ownership, sector
of location, and start-up year, which allows us to calculate the number of foreign firms
entering the Irish economy in every year. A firm is classified as being foreign-owned if 50
percent or more of its shares are held by foreign owners. Using these data we are able to
calculate the number of entrants from the US and the UK per year for 27 manufacturing
sectors.4
In addition we supplement our employment survey with information from the Forfás Irish
Economy Expenditure Survey and the Forfás Survey of Research and Development in
Industry in order to calculate some of the explanatory variables used in the econometric
analysis. The former database provides data on input-sourcing behaviour for a sample of
large firms (greater than 30 employees from 1983 and greater than 20 employees from 1990
onwards) since 1983. The latter survey has been undertaken since 1986 and provides data
on the population of R&D performers with ten or more employees in the manufacturing
sector. The use of these two datasets means that in our econometric analysis below we are
constrained to analysing the period 1986 to 1996.
Table 2 provides some descriptive statistics relating to the number of US and UK firms
based in Ireland. As indicated earlier these firms would appear to represent two different
types of foreign firms in Ireland. It is notable that the total number of US firms has grown
by 144 percent, while the number of UK firms declined by 54 percent over the period 1973
to 1996. One can also see differences in the sectoral location of firms. While US firms are
much more concentrated in modern high-tech sectors (such as machinery, transport and
scientific equipment), UK firms are more concentrated in low-tech traditional sectors (e.g.,
food, textiles & leather, paper & printing). These different characteristics may suggest that
US and UK firms also follow different strategies in their location decision, a question that
we investigate further below.
[Table 2]
Machinery and Equipment, Transport Equipment and Electrical and Optical Equipment.
4
The sectoral classification is based on a comparison of ISIC and NACE sectoral classification. We linked
these two sectoral classifications in order to be able to link the data used in the analysis below.
5
3.
Econometric Model and Methodology
In order to investigate in more detail whether the entry of new foreign firms in Ireland is
related to the presence of efficiency agglomerations and demonstration effects, we need to
model the location decision of foreign firms.
We, therefore, postulate the following
empirical model which relates the number of foreign entrants of nationality group f (f = US,
UK), nf, in sector j at time t to a number of explanatory variables,
n jtf = β 0 + β1 A jt + β 2 D jt + β 3 X t + u jt
(1)
where A and D are efficiency agglomeration and demonstration effects respectively, and X
denotes other observables which may affect nf. Of course, A and D are unobservable and
we need to find appropriate proxies. What is observable, however, is the total spatial
localisation of foreign firms, LOC, which is due to firms clustering either to benefit from
efficiency agglomeration benefits and/or from demonstration effects,
LOC = A + D
(2)
In other words, we can observe the composite LOC but the two components A and D are
unobservable. In order to find appropriate proxies for A and D we therefore suggest to run
the following regression
~
LOC jt = α 0 + α1 A jt + v jt
(3)
~
where A is a vector of variables to proxy the importance of efficiency agglomeration
benefits.
Using the result of this regression we estimate the predicted value of the
regression LOˆ C and the residual v̂ . We take the former as a proxy for A and the latter as a
proxy for D in the initial regression (1).
Proxies are, of course, always only approximations for the unobservable variable of interest,
and we are cautious to point out that particularly v̂ is a less-than-perfect proxy for the
unobservable D. While we focus on efficiency agglomeration and demonstration effects as
reasons for the spatial clustering of firms, traditional trade theory would suggest that
localisation of industries emerge purely due to endowment reasons. In a nutshell, firms will
locate in regions with favourable factor endowments. Our proxy for D therefore likely
includes demonstration effects as well as endowment driven agglomerations. We control
for this in the estimation of the entry model (equation (1)) by including a proxy for
endowment driven localisation, similar to Head et al. (1995). Our proxy for D should,
6
therefore, in the estimation of equation (1) provide an indication of demonstration effects,
after controlling for endowment effects.
We measure the extent of localisation of foreign firms (LOC) in sector j using (i) the total
number of foreign firms present in the sector (as in Head et al. 1995, 2000) and (ii) the (log
of) total employment stock in foreign firms in sector j. While Head et al. use the number of
firms as proxies for localisation variables on the right-hand side we note that this measure
does not allow them to distinguish between large and small firms.
We feel such a
distinction is important for at least two reasons. First the scale of other firms' investments
(reflected in employment size in Ireland), rather than the actual number of investments,
could represent an important demonstration effect. Secondly, attracting large multinational
firms (or "flagship projects") is likely to be more important than attracting smaller firms in
order for demonstration effects to emerge.
Replacing the number of firms by the
employment variable on the right-hand side takes account of the first effect and, if large
multinationals are likely to employ higher numbers in their Irish operations than are smaller
multinationals, should take account of the second point also.5
~
A includes three variables with which we try to proxy the three efficiency reasons why
firms may agglomerate, as discussed above.
Firstly, we include the R&D intensity
(spillover) in a sector in order to proxy for potential knowledge spillovers between firms.
This variable takes into account that spillovers arise when one firm’s innovative activity
leads to new ideas and an enhancement of innovative activity in a second firm without the
second firm having to compensate the other inventor.6
We measure a sector’s R&D
intensity as the proportion of total employment in R&D active firms. We would expect a
positive sign on the coefficient of spillover.
A measure of excess job turnover (turnover) is the second variable included. This variable
should control for the effect of the presence of thick labour markets in an agglomeration. If
there are thick markets for specialised labour adjustment costs can be presumed to be low,
as labour can move easily and hiring and firing costs are low. In such an environment,
workers tend to move more frequently between jobs, thus providing a readily accessible
5
Support for the latter hypothesis emerges from the database described in Pavelin (2000), on the 300 or so
leading firms in the EU (i.e. the top five firms in each 3-digit industry). These data display a positive
correlation between foreign firms' relative size rankings in their sector of activity in Ireland and in the overall
EU (though size in this database refers to production rather than employment).
7
common labour market pool for existing and potential firms within the sector. Hence we
choose to calculate the measure as the intra-industry job turnover in excess of inter-sectoral
employment shifts, as suggested by Davis and Haltiwanger (1992). A large degree of job
turnover indicates low adjustment costs; we would, therefore, expect a positive relationship
between excess and the presence of efficiency agglomerations.
The third variable is a proxy for the presence of input-output linkages between firms (link).
We calculate it as total raw materials, intermediate inputs, and services sourced in the Irish
economy per employee in a sector. This allows for the fact that firms may agglomerate if
there are input-output linkages between customer and supplier firms. We would, thus,
expect a positive sign on this variable.
While equation (3) can be estimated using OLS, the dependent variable in equation (1) is a
discrete variable and we, therefore, need to employ a count data model to estimate it. The
standard method is to assume that the variable is generated by a Poisson distribution of the
form
(
nf
)
Pr ob(n jtf ) = µ jt e − µ n jtf !
(4)
where µ is the conditional mean of the distribution. It is then assumed that the expected
value of n, µ is log linearly dependent on some explanatory variables, and parameter
estimates of these variables can be obtained using maximum likelihood techniques. The
Poisson model imposes the restriction that the conditional mean of the dependent variable
equals its variance. If it is found that this restriction does not hold in the data, one may
employ a negative binomial distribution, which allows for “overdispersion” in the data, i.e.,
the variance of the dependent variable is allowed to exceed the mean. In our econometric
analysis below we test for this restriction and find that, in all cases, we cannot reject the
assumption that the variance equals its mean.7 We therefore take the estimates generated by
the Poisson estimation as being appropriate.
6
This is the definition of knowledge spillovers used by, for example, Branstetter (2000).
See Bloningen (1997) and Coughlin and Segev (2000) for recent discussions of the Poisson and negative
binomial models, and applications in the analysis of location decisions of FDI. Note that, strictly speaking, the
Poisson specification is a special case of the negative binomial in which the overdispersion parameter is equal
to zero. Preliminary regressions, in which we estimated equation (1) using negative binomial regression
produced results which are quantitatively and qualitatively similar, suggesting that our choice of estimation
technique does not bias our results.
7
8
Apart from the proxies for A and D, equation (1) includes a number of other control
variables. These are the following:8
Relative cost competitiveness of Ireland as a host country (comp). We include a measure of
Ireland’s relative cost competitiveness as a host country in the EU, similar to Barrell and
Pain (1999). For UK companies it is straightforward to calculate such a variable. UK firms
are likely to decide between locating in Ireland or remaining in the home country and
therefore, relative labour costs between Ireland and the UK may be an adequate measure of
cost competitiveness. US companies, by contrast, may be assumed to search for alternative
locations in the EU in order to serve the EU market. Since Ireland and the UK share a
number of common characteristics, such as a common language and similar culture, similar
labour market institutions, similar location, it may appear reasonable to assume that Ireland
competes primarily with the UK for investments from the US. Hence, we measure Ireland's
relative cost competitiveness also as relative labour costs between Ireland and the UK. This
is calculated as the ratio of real wages and salaries per employee in sector j in Ireland
relative to the UK, converted to a common currency. To construct the variable we use data
from the Irish Census of Industrial Production and the UK Census of Production.
GDP growth in source country f (gdpgf). This variable is intended to control for the foreign
supply of FDI, as in Bloningen (1987). The assumption is that growth in the source country
is likely to generate a greater supply of FDI. Data for this variable were obtained from the
Bureau of Economic Analysis in the US Department of Commerce for US GDP, and
Eurostat for the UK data.
Size of the sector (size). The rationale for including this variable is to control for the fact
that one would expect larger numbers of entrants in large sectors. Since the Irish market is
very small, and foreign firms mainly locate in Ireland to service the larger European market
(see Barry and Bradley, 1997) we measure this variable as the size of the sector in the EU.9
The variable is calculated in terms of employment size, using data available from the
UNIDO database.
8
In preliminary regressions we also included sectoral dummies in the estimation of equation (1) to control for
sector-specific fixed effects. Tests indicated, however, that we cannot reject the hypothesis that all
coefficients of the sectoral dummies were jointly equal to zero. We conclude therefore that the specification
without sectoral dummies, the results of which are reported below, is preferable.
9
To be precise, the variable is calculated using data for France, Germany, Italy, the Netherlands and the UK
due to data constraints.
9
Ireland’s comparative advantage (adv). This variable is included to capture the effect of
endowments on industry location, as discussed above. All other things equal, foreign firms
should be expected to locate where factor endowments are favourable. We postulate that
the sectoral distribution of Irish-owned firms reflects this kind of information. We calculate
an employment specialisation index as the ratio of the share of sector j employment in Irishowned firms over total manufacturing employment in Irish-owned firms in Ireland relative
to the same share for the whole EU (including domestic and foreign-owned firms),

advt =  EtjIRL

∑E
j
IRL
tj




 EU
 Etj


∑E
j
EU
tj




(5)
where EjIRL is employment in Irish-owned firms in sector j in Ireland, and EjEU is
employment in both domestic and foreign-owned firms in sector j in the total EU, using the
same datasource as used for the calculation of sectoral size.10
4.
Econometric Results
As pointed out above, firms from the US and the UK appear to represent two very different
categories of foreign entrants in Ireland. Due to these differences, we may expect the
behaviour and location decisions of entrants of these two nationality groups to be different
also. We, therefore, present the results for analysing the effects of efficiency agglomeration
and demonstration effects on the entry of firms from these two nationality groups
separately.
10
This measure of comparative advantage (see also Barry and Hannan, 1996) also allows us to take into
account that foreign firms may be attracted to a sector simply because Ireland has a traditional comparative
advantage in that sector. Milner and Pentecost (1996) and Driffield and Munday (2000) show that revealed
comparative advantage has been an important determinant of inward FDI in the UK.
10
4.1 US entrants
Table 3 presents the results for the estimation of equation (3). Columns (1) and (2) relate to
estimations using the number of firms as measures of LOC (i.e., localisation of firms)
columns (3) and (4) report results based on estimations using the stock of employment as
the measure of localisation. Also note that in columns (1) and (3) the localisation measure
is calculated using all foreign firms, while (2) and (4) are based on these measures being
calculated for US firms only. This distinction should allow us to investigate whether the
benefits from efficiency agglomerations and demonstration effects emanate from all foreign
firms, or from firms of the same nationality only.
Inspection of the results shows that all estimated coefficients are of the right sign. The
measures of knowledge spillovers and labour turnover are statistically significant in all
cases, while the measure of linkages is only statistically significant in one case.
[Table 3]
Tables 4 and 5 then present the results of estimating equation (1), i.e., analysing the effect
of efficiency agglomerations and demonstration effects on the entry of US firms. The
results in Table 4 are based on measuring the localisation of firms using firm numbers,
while Table 5 is based on localisation of firms calculated as total employment stock. In
both tables, the results reported in columns (1) to (3) relate to estimations using the spatial
localisation of all foreign firms as a basis for calculating A and D, while columns (4) to (6)
present results for the localisation of US firms only. We also decomposed the data for all
manufacturing firms into groups of firms in high-tech and low-tech tech sectors to obtain
more homogenous comparison groups. Columns (2) and (5) show results using data on
firms in high-tech sectors only, while estimation results in columns (3) and (6) relate to
low-tech sectors only.11
We find in Table 4 that there is empirical evidence to suggest that, after controlling for
possible endowment effects and other factors, both efficiency agglomeration and
demonstration effects are important determinants for attracting new US firms.
11
The classification of sectors into high tech and low tech is based on an OECD classification as used by
Kearns and Ruane (2001). Accordingly, high tech sectors are Aerospace, Computers & Office machinery,
11
Comparisons of the sizes of the coefficients shows that, for high tech sectors, the coefficient
on A is larger than that on D, implying that the efficiency agglomeration effect appears to be
larger than the demonstration effect. Such a difference is not observable for low tech
sectors, however.
Furthermore, we find evidence for efficiency agglomerations and
demonstration effects when we examine the localisation of all foreign firms as well as when
looking at US firms only although the coefficients in the latter case are consistently higher.
This may suggest that both effects emanate more strongly from firms of the same
nationality as the entrant.
As regards to the other control variables included in the empirical model, we find
statistically significant evidence that US entry in the high tech sector decreases as Ireland’s
relative cost competitiveness vis-à-vis the UK worsens. There is no such evidence for the
low tech sector, however. This suggests that US entrants in the high tech sectors are
particularly likely to respond negatively to increases in Irish labour costs relative to the UK.
The positive and statistically significant (in three out of six cases) coefficient on adv
suggests that endowment effects, in addition to efficiency agglomerations and
demonstration effects, are also important for the location decisions of US entrants. As
theory would predict, US entrants are more likely to locate in sectors in which Ireland has
favourable factor endowments.
The results on the other two control variables are
statistically insignificant indicating that they do not appear to have any impact on the entry
of US firms.
[Table 4]
Using employment stock rather than firm numbers as a basis for our proxies A and D
produces the results reported in Table 5. In terms of the coefficients on A and D we find no
major changes to the results in Table 4; both are statistically significant, and the efficiency
agglomeration effect appears to dominate the demonstration effect.
One should note,
however, that now the coefficient on A is also higher than that on D in the low tech sector.
The measure of Ireland’s cost competitiveness is still negative for the high tech sector,
although it is now only statistically significant (at the ten percent level) in one case. The
Electronics & Communications, Pharmaceuticals, Scientific Instruments, Electrical Machinery, Motor
Vehicles, Chemicals, Non-electrical Machinery.
12
measure of comparative advantage shows also different results.
It is negative and
statistically significant in one case which, if taken at face value, would imply a negative
correlation between endowments and industry location. This is clearly contrary to what
theory would predict.
[Table 5]
4.2 UK entrants
The results for the entry of new UK firms into Ireland show somewhat different results on
the importance of efficiency agglomerations and demonstration effects. While the results of
estimating equation (3), reported in Table 6, are fairly similar to the results we obtained for
the US there are a number of differences apparent when inspecting the results of estimating
equation (1). These latter results are reported in Tables 7 and 8.
[Table 6]
When examining the localisation of all foreign firms (columns (1) – (3)) we find strong
evidence for positive effects emanating from demonstration effects, while efficiency
agglomeration effects only seem to matter for firms in low tech sectors. Furthermore,
limiting ourselves to the effects of localisation of UK firms shows evidence for
demonstration effects for firms entering in low tech industries, but not for efficiency
agglomerations. There are no such effects apparent for high tech industries. This may
suggest that efficiency agglomeration and demonstration effects originate mainly from firms
of other nationalities, while UK firms are not that important for the creation of efficiency
agglomeration and demonstration effects.
When looking at employment, rather than firm numbers, to proxy firm localisation (Table
8) we find consistently positive and statistically significant evidence for demonstration
effects. However, evidence of efficiency agglomeration effects is only to be found in the
case of UK entry in low tech industries. This strengthens the findings in Table 7 that
efficiency agglomerations do not seem to matter for UK entrants, in particular in high tech
industries. It also shows, however, that a distinction between measuring the localisation of
firms in terms of firm numbers or employment stock yields slightly different results.
[Tables 7 and 8]
13
5.
Summary and Conclusions
It has been established in the literature that the localisation of firms can be an important
factor in attracting new foreign direct investment into a host country. What has been
missing in the literature thus far, however, is an investigation into the reasons why industry
clusters attract firms. On the one hand, new foreign firms may be attracted because they
can increase their efficiency by locating close to other firms; this is the reason for
agglomerations frequently postulated in the new economic geography literature. Apart from
such "efficiency agglomerations” firms might also be attracted by the presence of existing
firms because of demonstration effects, whereby existing firms send signals to new
investors as to the reliability of the host country.
In this paper we try to disentangle these two reasons for industry localisations, by
examining the case of US and UK firms locating in the Irish economy. We calculate
proxies for “efficiency agglomerations” and “demonstration effects” and include these
proxies in an empirical model of the location decision of firms. For US firms, we find that
both efficiency agglomeration and demonstration effects are important determinants of
entry. For UK firms, however, the evidence is not as clear-cut. While demonstration
effects are important for the location of UK entrants there is no evidence to suggest that
efficiency agglomerations matter as well.
Our analysis underlines the different
characteristics of US and UK firms in Ireland and shows that these differences are also
reflected in the location decisions of firms from the two nationality groups.
On a more general level our distinction of efficiency agglomeration and demonstration
effects also suggests policy implications.
If firms are attracted by the former, the
government can assist the build up of such agglomerations through educational policies,
support of sub-supply industries etc. On the other hand, if firms are only attracted because
of demonstration effects, it is important from an economic development point of view to
attract a significant number of firms into the host country which are able to signal to other
firms the reliability of the host country. As the evidence suggests, Ireland seems to have
been able to attract such “flagship projects” at an early stage of development and now reaps
the benefits.
14
References
Barrell, Ray and Nigel Pain (1999), “Domestic Institutions, Agglomerations and Foreign
Direct Investment in Europe”, European Economic Review, Vol. 43, pp. 925-934.
Barry, Frank and John Bradley (1997), “FDI and Trade: The Irish Host-Country
Experience”, Economic Journal, Vol. 107, pp. 1798-1811.
Barry, Frank and Aoife Hannan (1996) "On Comparative and Absolute Advantage: FDI and
the Sectoral and Spatial Effects of Market Integration", Working Paper 96/19,
Centre for Economic Research, University College Dublin.
Bloningen, Bruce A. (1997), “Firm-specific Assets and the Link between Exchange Rates
and Foreign Direct Investment”, American Economic Review, Vol. 87, pp. 447-465.
Branstetter, Lee (2000), “Is Foreign Direct Investment a Channel of Knowledge Spillovers?
Evidence from Japan’s FDI in the United States”, NBER Working Paper No. 8015.
Braunerhjelm, Pontus and Roger Svensson (1996), “Host Country Characteristics and
Agglomeration in Foreign Direct Investment”, Applied Economics, Vol. 28, pp. 833840.
Braunerhjelm, Pontus and Roger Svensson (1998), “Agglomeration in the Geographical
Location of Swedish MNFs”, in Pontus Braunerhjelm and Karolina Ekholm (eds.),
The Geography of Multinational Firms (Kluwer Academic Publishers, Dordrecht),
pp. 99-115.
Coughlin, Cletus C. and Eran Segev (2000), “Location Determinants of New ForeignOwned Manufacturing Plants”, Journal of Regional Science, Vol. 40, pp. 323-351.
Davis, Steven J. and Haltiwanger, John (1992), "Gross Job Creation, Gross Job Destruction
and Job Reallocation", Quarterly Journal of Economics, Vol. 107, pp. 819-864.
DeCoster, Gregory P. and William C. Strange (1993), “Spurious Agglomeration”, Journal
of Urban Economics, Vol. 33, pp. 273-304.
Driffield, Nigel and Max Munday (2000), “Industrial Performance, Agglomeration, and
Foreign Manufacturing Investment in the UK”, Journal of International Business
Studies, Vol. 31, pp. 21-37.
Fujita, Masahisa and Jacques-François Thisse (1996), “Economics of Agglomeration”,
Journal of the Japanese and International Economies, Vol. 10, pp. 339-378.
Head, Keith; John Ries and Deborah Swenson (1995), “Agglomeration benefits and
location choice: Evidence from Japanese manufacturing investments in the United
States”, Journal of International Economics, Vol. 38, pp. 223-247.
Head, Keith C.; John C. Ries and Deborah L. Swenson (2000), “Attracting Foreign
Manufacturing: Investment Promotion and Agglomeration”, Regional Science and
Urban Economics, Vol. 29, pp. 197-218.
Kearns, Allan and Frances Ruane (2001), “The Tangible Contribution of R&D Spending by
Foreign-Owned Plants to a Host Region: A Plant Level Study of the Irish
Manufacturing Sector (1980-1996)”, Research Policy, Vol. 30, pp. 227-244.
Krugman, Paul R. and Anthony J. Venables (1995), “Globalisation and the Inequality of
Nations”, Quarterly Journal of Economics, Vol. CX, pp. 857-880.
Krugman, Paul R. and Anthony J. Venables (1996), “Integration, Specialization, and
Adjustment”, European Economic Review, Vol. 40, pp. 959-967.
15
Marshall, Alfred (1920), Principles of Economics. 8th Edition. London: Macmillan.
Milner, Chris and Eric Pentecost (1996), “Locational Advantage and US Foreign Direct
Investment in UK Manufacturing”, Applied Economics, Vol. 28, pp. 605-615.
O'Malley, Eoin (1995), An Analysis of Secondary Employment Associated with
Manufacturing Industry. General Research Series Paper No. 167. Dublin: Economic
and Social Research Institute.
Ottaviano, Gianmarco I.P. and Diego Puga (1998), “Agglomeration in the Global Economy:
A Survey of the ‘New Economic Geography’”, The World Economy, Vol. 21, pp.
707-731.
Pavelin, Stephen (2000) "The Geographical Diversification of Leading Firms in the EU",
Working Paper 00/15, Centre for Economic Research, University College Dublin.
Wheeler, David and Ashoka Mody (1992), “International investment location decisions:
The case of U.S. firms”, Journal of International Economics, Vol. 33, pp. 57-76.
White, Padraic (2000), “The Muscles of the Celtic Tiger: The IDA’s Winning Sectors”, in
Ray MacSharry and Padraic A. White (eds.), The Making of the Celtic Tiger: The
Inside Story of Ireland’s Boom Economy (Mercier Press, Cork), pp. 272-308.
16
Tables
Table 1: Comparison between US- and UK-owned firms in Ireland (1998)
Firm
Ownership
US
UK
Number
of firms
295
122
Proportion of
manufacturing
employment
27%
5%
Value-added
per employee
(£000)
258.4
113.5
Proportion
of own gross
output
exported
97%
58%
High-tech sectors
% of
firms
% of
empl
% of
net Y
62%
39%
73%
28%
69%
28%
Source: Census of Industrial Production
Table 2: Percentage of firms by sector, US and UK
Sector
US
Food, Drink & Tobacco
Textiles & Leather
Wood & Furniture
Paper & Printing
Chemicals
Rubber & Plastics
Non-metallic minerals
Metals
Machinery
Transport Equipment
Scientific Equipment
Other manufacturing
Total Number
UK
1973
1985
1996
1973
1985
1996
9.8%
7.5%
1.5%
3.8%
17.3%
4.5%
3.0%
15.8%
16.5%
5.3%
11.3%
3.8%
6.6%
6.9%
1.0%
3.1%
12.5%
5.2%
2.1%
8.7%
33.9%
4.5%
13.1%
2.4%
6.2%
4.9%
1.5%
3.7%
14.5%
5.2%
0.9%
6.5%
32.7%
7.1%
14.2%
2.5%
18.2%
23.2%
2.5%
4.7%
10.0%
5.0%
10.7%
12.2%
6.3%
3.1%
1.6%
2.5%
19.0%
17.6%
1.0%
2.9%
14.3%
5.7%
13.8%
11.0%
9.0%
2.9%
1.9%
1.0%
21.9%
9.6%
0.0%
4.1%
17.8%
11.6%
10.3%
11.6%
9.6%
1.4%
1.4%
0.7%
133
289
324
319
210
146
Source: Own calculations using Forfás Employment Survey data
17
Table 3: Auxiliary regression for US firms
OLS regression
spillover
turnover
link
constant
Observations
F(ßi=0)
R2
(1)
(2)
(3)
(4)
Foreign firm
numbers
US firm
numbers
Foreign
employment
US
employment
56.271
(10.392)***
255.994
(40.934)***
0.262
(0.075)***
-30.859
(8.413)***
27.773
(4.828)***
107.392
(19.017)***
0.053
(0.035)
-16.364
(3.908)***
1.725
(0.604)***
13.598
(2.381)***
0.003
(0.004)
-0.778
(0.489)
1.442
(0.493)***
8.588
(1.943)***
0.002
(0.004)
-0.949
(0.399)**
250
22.03***
0.20
250
18.82***
0.18
250
12.17***
0.12
250
8.20***
0.08
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.
Table 4: Results of entry regression for US firms: firm numbers
Poisson regression
(1)
At
Dt
compt
gdpgUSt
advt
sizet
constant
Observations
LR(ßi=0)
R2
LOC: all foreign firms
(2)
(3)
(4)
LOC: only US firms
(5)
(6)
All
High tech
Low tech
All
High tech
Low tech
0.036
(0.006)***
0.020
(0.003)***
-0.722
(0.643)
4.410
(7.132)
-0.234
(0.144)
-0.2e-07
(1.3e-07)
-1.559
(0.714)**
0.036
(0.009)***
0.007
(0.004)*
-2.641
(1.058)***
10.887
(8.402)
4.811
(2.876)*
0.8e-07
(2.1e-07)
-1.117
(1.798)
0.031
(0.018)*
0.025
(0.017)
-0.453
(1.115)
11.283
(15.261)
0.364
(0.223)
-3.9e-07
(5.8e-07)
-2.724
(1.109)**
0.071
(0.016)***
0.047
(0.006)***
-0.507
(0.643)
3.732
(7.188)
0.062
(0.154)
-0.4e-07
(1.3e-07)
-1.639
(0.722)**
0.060
(0.026)**
0.028
(0.008)***
-2.367
(1.061)**
7.544
(8.489)
4.461
(2.579)*
0.6e-07
(1.8e-07)
-0.889
(1.627)
0.234
(0.062)***
0.246
(0.067)***
0.305
(1.137)
9.427
(15.713)
0.512
(0.252)**
-16.4e-07
(5.9e-07)***
-3.373
(1.111)***
250
212.59***
0.36
60
76.58***
0.32
190
14.97***
0.08
250
234.90***
0.40
60
84.32***
0.35
190
27.23***
0.14
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.
18
Table 5: Results of entry regression for US firms: employment
Poisson regression
(1)
At
Dt
compt
gdpgUSt
advt
sizet
constant
Observations
LR(ßi=0)
R2
LOC: all foreign firms
(2)
(3)
(4)
LOC: only US firms
(5)
(6)
All
High tech
Low tech
All
High tech
Low tech
0.873
(0.146)***
0.626
(0.063)***
0.119
(0.734)
3.393
(6.750)
-0.396
(0.139)***
-0.5e-07
(1.2e-07)
-2.797
(0.825)***
0.774
(0.186)***
0.462
(0.106)***
-1.518
(1.101)
6.352
(8.080)
2.236
(2.770)
-1.6e-07
(2.0e-07)
-1.292
(1.636)
0.923
(0.310)***
0.552
(0.112)***
0.170
(1.266)
10.048
(14.366)
0.278
(0.248)
-3.5e-07
(3.4e-07)
-4.136
(1.313)***
0.805
(0.235)***
0.647
(0.055)***
-0.336
(0.702)
6.714
(6.854)
-0.088
(0.143)
-0.2e-07
(1.2e-07)
-1.896
(0.742)***
0.703
(0.315)**
0.417
(0.079)***
-1.678
(1.043)*
12.734
(8.010)
2.414
(2.602)
-0.5e-07
(1.8e-07)
-0.720
(1.593)
1.501
(0.545)***
0.894
(0.119)***
-0.335
(1.285)
-12.622
(15.950)
0.117
(0.273)
-1.2e-07
(3.0e-07)
-2.675
(1.171)**
250
262.63***
0.45
60
93.34***
0.39
190
40.93***
0.22
250
312.28***
0.53
60
105.54***
0.44
190
77.04
0.40
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.
19
Table 6: Auxiliary regression for UK firms
OLS regression
spillover
turnover
link
constant
Observations
F(ßi=0)
R2
(1)
(2)
(3)
(4)
Foreign firm
numbers
UK firm
numbers
Foreign
employment
UK
employment
56.271
(10.392)***
255.994
(40.934)***
0.262
(0.075)***
-30.859
(8.413)***
5.317
(2.188)**
20.342
(8.620)**
0.093
(0.016)***
-0.787
(1.771)
1.725
(0.604)***
13.598
(2.381)***
0.003
(0.004)
-0.778
(0.489)
0.428
(0.374)
3.840
(1.475)***
-0.003
(0.003)
-0.080
(0.303)
250
22.03***
0.20
250
13.69***
0.13
250
12.17***
0.12
250
3.03***
0.02
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.
Table 7: Results of entry regression for UK firms: firm numbers
Poisson regression
(1)
At
Dt
compt
gdpgUKt
advt
sizet
constant
Observations
LR(ßi=0)
R2
LOC: all foreign firms
(2)
(3)
(4)
LOC: only UK firms
(5)
(6)
All
High tech
Low tech
All
High tech
Low tech
0.015
(0.009)
0.017
(0.004)***
-0.045
(0.784)
6.517
(5.926)
0.284
(0.144)**
0.6e-07
(2.0e-07)
-2.599
(0.810)***
-0.013
(0.027)
0.023
(0.010)**
-1.642
(1.861)
13.537
(11.805)
-3.826
(5.968)
-2.9e-07
(4.6e-07)
1.838
(3.841)
0.070
(0.019)***
0.077
(0.018)***
0.739
(0.981)
-2.267
(7.040)
0.096
(0.198)
-13.1e-07
(5.3e-07)**
-3.222
(0.977)
0.064
(0.056)
0.079
(0.017)***
-0.142
(0.865)
5.204
(5.950)
-0.010
(0.156)
4.0e-07
(1.6e-07)***
-2.433
(0.833)***
0.152
(0.175)
0.041
(0.043)
-2.233
(2.036)
18.925
(11.685)
2.240
(5.529)
2.2e-07
(3.8e-07)
-1.774
(3.455)
0.047
(0.070)
0.096
(0.031)***
0.494
(1.141)
-0.013
(7.097)
-0.003
(0.204)
3.4e-07
(3.1e-07)
-2.858
(1/050)***
250
40.81***
0.12
60
13.26**
0.15
190
50.48***
0.21
250
41.85***
0.13
60
8.54
0.10
190
35.89
0.15
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.
20
Table 8: Results of entry regression for UK firms: employment
Poisson regression
(1)
At
Dt
compt
gdpgUKt
advt
sizet
constant
Observations
LR(ßi=0)
R2
LOC: all foreign firms
(2)
(3)
(4)
LOC: only UK firms
(5)
(6)
All
High tech
Low tech
All
High tech
Low tech
0.595
(0.221)***
0.501
(0.082)***
0.579
(0.866)
5.459
(5.699)
0.143
(0.136)
-0.7e-07
(1.7e-07)
-3.677
(0.980)***
0.079
(0.480)
0.426
(0.218)**
-0.787
(1.951)
15.853
(11.360)
0.877
(5.059)
-0.5e-07
(3.9e-07)
-1.900
(3.313)
0.908
(0.273)***
0.600
(0.103)***
1.497
(1.067)
0.229
(6.681)
-0.327
(0.202)
3.8e-07
(2.5e-07)
-4.844
(1.156)***
1.606
(0.851)**
0.706
(0.069)***
1.162
(0.888)
2.575
(5.938)
0.166
(0.166)
4.7e-07
(1.5e-07)***
-4.811
(1.235)***
3.148
(2.519)
0.869
(0.203)***
0.539
(2.183)
2.786
(12.581)
4.547
(5.178)
2.1e-07
(3.2e-07)
-6.469
(4.656)
1.878
(1.005)*
0.660
(0.079)***
1.839
(1.165)
-0.134
(7.229)
0.141
(0.227)
6.8e-07
(2.5e-07)***
-5.749
(1.504)***
250
68.31***
0.21
60
11.77*
0.14
190
70.15***
0.29
250
129.12***
0.39
60
30.63***
0.35
190
103.08***
0.43
Notes: Standard error in parentheses.
*** = statistically significant at 1 per cent, ** at 5 per cent, * at 10 per cent level.